mpfaffenberger/code_puppy

Agentic AI for writing code

What it solves

Code Puppy is an AI-powered code generation agent designed to replace expensive IDEs like Cursor or Windsurf by providing a flexible, open-source alternative for understanding programming tasks, generating high-quality code, and explaining its reasoning.

How it works

It operates as a command-line agent that can be configured with various LLM providers. It uses a system of specialized agents (both built-in Python agents and user-defined JSON agents) that can access a suite of tools for file management, shell command execution, and text search. It also supports Model Context Protocol (MCP) servers for external tools and durable execution via DBOS to ensure agent runs survive crashes.

Who it’s for

Developers who want a powerful, AI-driven coding assistant that is not tied to a specific IDE, supports a wide range of models (via models.dev integration), and offers deep customization of agent behaviors and rules.

Highlights

  • Extensive Model Support: Integrates with models.dev to access over 1,000 models from 65+ providers.
  • Customizable Agents: Create specialized assistants using a guided "Agent Creator" or manual JSON configurations.
  • Durable Execution: Optional DBOS integration allows users to resume long agent runs after a crash.
  • Agent Rules: Supports AGENTS.md files for project-specific coding standards and behavioral guidelines.
  • Round Robin Distribution: Distributes requests across multiple API keys to bypass rate limits.
  • Tool Suite: Built-in tools for reading, writing, and searching files, as well as executing shell commands.

Related

  • Project
  • Project
  • Project
  • Project
  • Project